Making sense of the polls

Making sense of the polls

Age Breakdowns show Huge Differences

Morgan’s latest poll from last weekend had a sample of over 4,500. Under the previous election’s preferences, the two party result was 52-48 to the Coalition; this was very much in line with the consensus. Morgan has been giving detailed demographic breakdowns, and he uses six age categories, while other pollsters use only three or four. The table below shows the primary votes for Labor, the Coalition and the Greens among the various age categories. Note that the two party columns use Morgan’s respondent-allocated preferences, except for the 65+ category, where Morgan’s 55-45 to the Coalition was clearly a typo, since the Coalition has 57% of the primary vote in that category. I’ve estimated this as 61-39 to the Coalition using the previous election’s preferences. The sample size for each age category is also provided.

age diff.

Labor and the Greens dominate the 18-24 and 25-34 categories, with the Coalition primary vote at 35% or below. The Coalition primary is ahead of Labor’s in the 35-49 category, but Labor still wins this category narrowly on Greens preferences. In the 50-64 category, the Coalition primary is just short of 50%, and it clearly wins that category. Senior citizens are very pro-Coalition; here, the Coalition has 57% of the primary vote, compared to Labor’s 33.5% and the Greens only have 3.5% in that category. This is the demographic group where the Coalition makes up for its huge deficit among young people.

There could be some good news for Labor in these age breakdowns. People who are leaning or undecided are more likely to be from the younger demographics, as seniors tend to be more rusted-on. If leaning and undecided voters break the way of their general demographic, Labor should do well. As Labor clearly needs to make up some ground to win the election, they need to win over late deciders.

Gender breakdowns are also given by Morgan, and Labor is ahead by 51.5-48.5 with women, but behind by 53.5-46.5 with men using respondent-allocated preferences. Both genders moved solidly to Labor when Rudd replaced Gillard, and Labor still does better with women than men.

A Ludicrous Poll in Rudd’s Seat of Griffith

A Lonergan robopoll shows Kevin Rudd losing his own seat of Griffith 52-48; this would represent a swing of over 10% to the Liberals in that seat. However, the latest state breakdowns from Essential, Newspoll and Morgan all show a small swing to Labor in Queensland, and the Poll Bludger’s Bludgertrack poll aggregate is now recording a swing of 1.5% to Labor in Queensland. The result in Griffith is thus completely out of alignment with what we would expect based on a small pro-Labor swing in Queensland. If Rudd is genuinely in danger, Labor would be facing a wipe-out in Queensland.

There are two reasons for why this poll is so aberrant. First, it is a robopoll, and only motivated voters are likely to respond to robopolls. Second, Lonergan weights its individual seat polls to the previous election’s results. As I previously noted, people misremember who they voted for at the last election. Weighting to past results seems especially problematic when the sitting member is high profile, and Rudd is Australia’s PM. Many people will say that they voted for Rudd when they actually voted for someone else at the 2010 election.